Custom Business Central AI Agents: Where to Start

Few phrases are more optimistic than the subject line “Quick question.”

It’s Monday morning, and this one comes from a customer who needs 12 of the same parts they ordered last time, plus six of the larger ones. And can you get them there by Friday?

Perfectly clear, right? Assuming, of course, you know which order was “last time,” what “the larger ones” are, whether you have them in stock, and whether Friday is remotely realistic.

This is exactly the kind of work that makes sales order processing an interesting use case for AI. The individual steps aren’t especially complicated. But moving from a customer’s email to an accurate sales document requires context, interpretation, and sometimes a question or two along the way.

In the first article in this series, I looked at how Business Central AI agents can take on routine work while people stay in control. In the second, I looked at Business Central AI agent ROI and what that recovered time could mean for the business.

Now, let’s take the next step: what happens when the process you want to improve is unique to your company?

The Business Central Sales Order Agent gives us a practical place to start. If an agent can work through a sales request like this, what other processes could custom Business Central AI agents take on?

Why sales order processing is a strong agent use case

That Monday-morning email is a good example of why sales order processing is such a natural fit for an agent.

The process usually follows a recognizable path: a customer sends a request, someone identifies the products and quantities, checks availability, fills in whatever information is missing, and turns the request into a quote or order.

In Dynamics 365 Business Central, much of the information needed to move that process forward is already there.

But “recognizable” doesn’t always mean predictable.

A customer might use an old product description instead of an item number. They may ask for a quantity that isn’t currently available. Or, as our Monday-morning customer so helpfully demonstrated, they may assume “the larger ones” is all the product identification anyone could possibly need.

The point here is that traditional automation works well when the path is clear: If this happens, do that.

But sales order intake often requires something more: interpreting the request, using the context in Business Central, and recognizing when a person needs to step in.

That’s where agentic AI workflows become interesting. The goal isn’t to remove the salesperson from the customer relationship. It’s to reduce the administrative work between “Can you get us these parts?” and a properly prepared sales document.

And that brings us to a useful real-world example: Microsoft’s Business Central Sales Order Agent.

What can the Business Central Sales Order Agent do?

The Business Central Sales Order Agent can monitor incoming customer emails and help move a sales request from the inbox toward a quote or order, while bringing an employee back into the process when review or judgment is needed.

Microsoft describes a process that starts when the agent identifies a potential customer

request in a designated mailbox. The agent can:

• verify the customer

• identify requested items

• check availability

• create a sales quote

• prepare an email response with the quote attached

If information is missing, the agent can continue the email exchange with the customer to help resolve it.

Think back to our Monday-morning customer. Instead of a salesperson manually searching for the previous order, identifying the parts, checking inventory, and building the quote, the agent can take on much of that routine legwork.

But this isn’t a case of turning over the sales inbox and hoping for the best.

Employees can review incoming requests, sales documents, and outgoing emails, depending on how the agent is configured. If information is missing or something needs to change, the agent can bring the task back to an employee for help.

That human involvement matters because not every customer request fits neatly into the standard process. Pricing, substitutions, credit issues, and special requests may still require someone who understands the customer and the business.

The value isn’t that AI replaces that judgment. It’s that your salespeople can spend less time getting a routine request from the inbox into Business Central and more time on the parts of the customer relationship that actually need them.

Moving beyond standard agents

The Sales Order Agent gives us something concrete to work with. You can see the process, identify the repetitive steps, and understand where an employee still needs to make a decision.

But your most time-consuming process may have nothing to do with sales orders. Maybe it’s inventory follow-up, customer onboarding, service coordination, purchasing checks, or getting reports ready.

That’s where custom Business Central AI agents start making a lot more sense.

Microsoft’s AI Development Toolkit for Business Central includes Agent Designer, which allows you to test what an agent might look like for one of your own processes. Just describe what you want an agent to accomplish, define its instructions and permissions, and test its behavior before moving toward a production solution.

That opens the door to Business Central agent design around processes unique to your operation.

But just because you can create an agent doesn’t mean every process needs one.

So how do you know which approach fits?

Standard agent, automation, or custom agent: Which approach fits the process?

Not every repetitive process needs custom ERP automation, and not every automation needs AI. The right choice depends on what the work actually requires.

If Microsoft already offers a standard agent that closely matches the process, start there. You’re working with a predefined capability designed for a familiar Business Central workflow rather than building something specifically for your company.

Traditional automation, including Power Automate, makes sense when the steps are predictable and the rules are clear. If this happens, do that. There’s not much interpretation required.

A custom agent starts to make more sense when the process isn’t quite so tidy. Maybe it involves several steps, information that needs to be interpreted in context, or company-specific rules and exceptions that can change what should happen next.

That's an important distinction. The goal isn’t to find somewhere to use AI. It’s to choose the simplest approach that can reliably handle the work.

If a standard agent or straightforward workflow solves the problem, great. But when neither quite fits, custom Business Central AI agents give you another option.

Before building one, though, there’s another question to answer: Is this truly a good process for an agent?

Custom Business Central AI agents: Your First Use Case

So, what makes a business process a good candidate for a custom AI agent? Start with the problem, not the technology.

Look for work that happens frequently, follows a recognizable series of steps, and has a clear outcome. The process should rely on dependable Business Central data, with clear rules around what the agent can do and when a person needs to step in.

That last part matters. Which decisions require approval? Which exceptions should go to an employee? What happens when the information the agent needs isn’t there?

And make sure there’s something you can measure. Maybe it’s faster response times, fewer manual steps, less rework, or simply giving employees back time for higher-value work.

Because the best first custom agent probably isn’t the most ambitious one. It’s the one that solves a real problem and gives you a clear way to tell whether it worked.

Where do you go from here?

The Sales Order Agent shows what’s possible when AI handles a familiar process without taking people out of the important decisions. Custom Business Central AI agents extend that idea to the processes unique to your business.

The place to start isn’t with the biggest idea. It’s with the right problem.

If you have a repetitive process that might be a good fit, reach out. We can help you determine whether an agent makes sense before you start building.

And whether you're exploring Business Central AI agents, evaluating a move from Dynamics GP, or simply trying to get more from Business Central, I host open Office Hours every Tuesday at 11:30 ET. Bring your questions and let's talk through what makes sense for your business.

About the Author

Chris Boling is a founding partner of Sandlapper Dynamics, where he helps businesses streamline operations, enhance productivity, and achieve strategic growth through Microsoft Dynamics 365. With over two decades of experience in the Dynamics community, Chris combines deep technical expertise with a customer-first approach to guide organizations through digital transformation.

His unique perspective, shaped by years as both a consultant and an end-user, enables Chris to deliver practical insights that bridge the gap between technology and business outcomes.

Chris brings authenticity, empathy, and a commitment to sustainable growth to every engagement.

You can reach Chris on LinkedIn.